The Economics of Happiness in Argentina

ISSN 0328-5715
ISSN 2524-955X
The Economics of Happiness in Argentina
Martin Tetaz•
Abstract
Since Easterlin famous paper, in 1974, many others tried either to refute its findings or
to produce additional explanations. Scholars have found relative income and adaptation
effects, threshold effects, age and gender effects, and so on. However few papers analyzed
the determinants of happiness, using data from Argentina. In this paper we exploit two
different data sets from Argentina. The main findings indicate a raise in happiness in
the last 10 years, although with regional differences. Income effects are significant but
not always present. The same applies to age effects. Neither gender nor the number of
children has any impact on happiness. Those very active at family and social life report
higher levels of happiness. Finally couples very active at their relationships are happier,
even when they are not as active in their sexual lives.
JEL CODE: D01, D03, D60
• CEDLAS-IIL-UNLP.
Palermo Business Review | Nº 7 | 2012|————————————————————————————————————————————————————| 41
Martin Tetaz
1. Introduction
In 1974, analyzing data on happiness for 19 countries, Richard Easterlin found a
puzzling result, known since then as “the Easterlin paradox”. In a nutshell, his main
finding was that although within a country there was a positive correlation between income
and happiness, “the association among countries was uncertain”. Moreover, for the only
country he had time series data –United States– “higher income was not systematically
accompanied by grater happiness”.
Since then an impresive amount of papers have tried to explain the issue (see Powdthavee
(2007); Veenhoven and Hagerty (2006) for reviews). For instance, Clark, Frijters and
Shields (2006) argue that if happiness depends on relative income considerations and
consumption, the first effect should not show up at aggregate levels and the second is not
likely to increases too much in developed countries. In the same direction Graham (2008)
states that “humans are on a hedonic treadmill” whereby happiness increases untill basic
needs are satisfied but then relative more than absolute income matters. The explanation
of why this effect takes place may have to do with an adaptation effect (Fray and Stutzer
(2002); Di Tella and MacCulloch (2008)).
Stevenson and Wolfers (2008) state that it is a matter of analyzing the issue with
appropriate data. Using a broad number of countries and making adjustments so as to
make the different surveys comparable, they find income coefficients of 0.3 and 0.36 for
within and between country respectivly (both statistically significant). They also obtained
a positive time series effect for most of the countries with the exception of the United
States. Taken all toghether they conclud that there is room for both, absolute and relative
income effects to account for people happiness.
Powdthavee (2010) uses an instrumental variable approach and find that when
controlling for endogeneity, not only income coefficients remains statistically significant,
but they actually doubles. In their attempts to replicate Easterlin’s findings, scholars have
came across other different effects, such as the “age effect” (Blanchflower and Oswald
2004); gender effects (Stevenson and Wolfers 2008); interaction effects between both
(Plagnol and Easterlin 2008); and many other socioeconomic effects (see Vennhoven
1994 for a review).
For example, Powdthavee (2012) analyzes the influences of changes on “the Big Five”
personality traits coming to the conclusion that they matter even more than socioeconomic
factors.
Aknin (et. al. 2012) focuses on the way money is spent rather than the actual levels of
income finding that pro social behaviors increases happiness more than money. Finally
Kahneman’s last book (2011) suggests that the key to happiness may not be income, but
the alternative allocations of personal time.
42|————————————————————————————————————————————————————|Palermo Business Review | Nº 7 | 2012
The Economics of Happiness in Argentina
Interesting as it may be, however, there are few studies on the economics of happiness
using data from Argentina. Giarrizzo (2008) in a tailored survey, carried out by Centro
de Economía Regional y Experimental (CERX) and Centro de Investigaciones en
Epistemología de las Ciencias Económicas (CIECE) finds that although 84% of the
respondents evaluated their economic situation as either “very bad”, “bad” or “regular”,
an overwhelming majority (73.5%) considered itself “happy” or “very happy”. However,
when they were asked to answer what factors would make them happier 81.5% of the
surveyed referred to economic related aspects, such as a higher income or a better job.
Cruces, Ham and Tetaz (2008) offer an analysis of quality of life and happiness at a
neighbourhood level. Both, quality of the neighbourhood and income had a positive
impact on happiness. They also found a positive correlation of some interesting variables,
such as the home quality, the satisfaction with friend, mental health and emotional life.
In addition, males were happier on average than women. Neither the number of children
nor the age had a statistically significant effect.
In this paper we attempt to analyze the micro determinants of happiness using two
different data sets from Argentina. The well known World Value Survey, covering five
waves and two waves of the 2011 Gallup – Universidad de Palermo survey.
Whereas the first data set allows us to study the relation between many socioeconomic
outcomes and happiness, the latter survey provides additional information on the perceived
determinants of subjective well being and it’s relation to people’s use of time.
The rest of the paper is organized in the following way. Section 2 presents some
stylized facts. Section 3 discusses the econometric strategy. In the next part of the paper
we show the econometric results of the World Value Survey data base. Section 5 explores
the relation between happiness and the satisfaction with the financial situation of the
households. Section 6 presents the findings of the Gallup-Universidad de Palermo surveys.
Section 7 concludes with some discussions and comments.
2. Stylized facts
To begin with, we would like to make a distinction between the two different questions
both, the World Value Survey and the Gallup – Universidad de Palermo survey ask in
order to elucidate respondents well being. The first question commonly referred to as
“Life satisfaction” is: “In general, taking all in, how satisfied are you with your life in
a 10 point scale, where 1 is dissatisfied and 10 is satisfied?”; The second question digs
straight forward into happiness asking: “In general, taking all in, would you say that you
are Very Happy; Quite Happy; Not Very Happy, or Not At All Happy?”.
Tables number 1, 2 and 3 show the relation between both questions, first in the five
waves of the World Value Survey, and then in the two waves of the Gallup-Universidad
de Palermo survey. Life satisfaction is in all cases the dependent variable.
Palermo Business Review | Nº 7 | 2012|————————————————————————————————————————————————————| 43
Martin Tetaz
Table 1 | The relation between Happiness and Life Satisfaction (OLS estimation)
Year 1984
Year 1991
Year 1995
Year 1999
Year 2006
very happy
1,422
0,730
1,272
0,677
0,981
t statistic
9,330
5,800
9,520
5,660
9,040
not very happy
-2,020
-1,186
-1,791
-1,893
-2,308
t statistic
-13,070
-6,760
-8,420
-9,310
-9,650
not happy al all
-4,068
-3,036
-4,152
-3,434
-3,787
t statistic
-6,090
-6,740
-8,860
-7,050
-6,310
_cons
6,982
7,341
6,926
7,477
7,745
t statistic
96,840
91,000
79,240
97,120
104,910
R2
0,290
0,180
0,250
0,180
0,280
Source; own calculations based on World Value Survey (All coefficients significant at 1%)
Table 2 | The relation between Happiness and Life Satisfaction (OLS estimation)
Coeficients
t statistic
very happy
0,916
10,300
not very happy
-2,333
-13,780
not happy al all
-4,485
-6,140
_cons
8,016
151,660
R2
0,430
Source; own calculations based on Gallup-Universidad de Palermo
(All coefficients significant at 1%)
Table 3 | The relation between Happiness and Life Satisfaction (OLS estimation)
Coeficients
t statistic
very happy
1,118
12,750
not very happy
-2,026
-10,340
not happy al all
-4,408
-7,430
_cons
7,911
160,900
R2
0,4
Source; Own calculations based on Gallup-Universidad de Palermo
(All coefficients significant at 1%)
44|————————————————————————————————————————————————————|Palermo Business Review | Nº 7 | 2012
The Economics of Happiness in Argentina
Since happiness variables are dichotomic, the “Quite happy” answer is always the
omitted response, so the coefficients can be interpreted as the change in life satisfaction
when the subject state changes from “Quite happy” to the other responses (i.e. very happy,
not very happy, not happy at all). All the regression show positive and statistically significant
coefficients ranging from 0,677 to 1,422 for “very happy”, and negative (and significant) ones
for “not very happy” and “not happy at all”.The goodness of fit of all models (R2) goes from
0,18 in the 1999 World Value Survey, to 0,43 in the first wave of the 2011 Gallup survey.
Due to the strong correlation with Happiness, and the ease of interpretation, we will
focus now on the Life satisfaction question, to see whether it has changed over time. To
start analyzing the evolution of Life Satisfaction, graph number 1 plot the average response
to the question “In general, taking all in, how satisfied are you with your life in a 10 point
scale, where 1 is dissatisfied and 10 is satisfied?” in all the surveys. It also display the 95%
confidence intervals, so the reader can easily tell when there is a significant difference
between two averages values (not intervals overlapping) and otherwise.
Graph 1 | “Taking all in how satisfied are you with your life”
Source; Own calculations based on World Value Survey and Gallup- Universidad de
Palermo (bars represent 95% confidence intervals)
At first sight there seems to be no statistically significant difference in happiness (life
satisfaction) between 1984 and 1995, neither is any noticeable discrepancy between 1991
and 1999. The contrast between 2006 and 2011 depends on the 2011 wave taken into
consideration. On the other hand, there is indeed an upward trend from 1995 to 2001.
A closer look at the five World Value Survey waves, however, suggest differences in
sampling strategies among them, rendering problematic the simple average comparisons.
To give the reader an idea of the differences, we show the kernel distribution of education
in three particular waves, as graph number two shows:
Palermo Business Review | Nº 7 | 2012|————————————————————————————————————————————————————| 45
Martin Tetaz
Graph 2 | Kernel distribution of education in three different waves
Source; Own calculations based on World Value Survey
Graph number 2 shows that, for example, the 2006 survey was administered to a more
educated sample. Furthermore, the heterogeneous sampling strategies over time can be
further confirmed checking out the kind of workers surveyed in the five waves. Table 4
highlights the differences.
Table 4 | Composition of worker’s skills in the different samples
Year 1984
Year 1991
Year 1995
Year 1999
Year 2006
“employer/manager of establishment
4%
9%
16%
10%
3%
“professional worker”
14%
9%
13%
1%
7%
“middle level non-manual office worker” 0%
12%
0%
0%
0%
“supervisory non manual -office worker” 33%
0%
2%
2%
4%
“junior level non manual”
0%
10%
0%
0%
0%
“non manual -office worker”
15%
0%
33%
25%
18%
“foreman and supervisor”
5%
6%
4%
1%
1%
“skilled manual”
15%
15%
9%
20%
19%
“semi-skilled manual worker”
12%
6%
4%
0%
8%
“unskilled manual”
1%
8%
13%
16%
19%
“farmer: has own farm”
0%
0%
0%
0%
1%
“agricultural worker”
0%
0%
0%
0%
1%
“member of armed forces”
0%
1%
1%
0%
1%
“never had a job”
0%
23%
2%
25%
16%
Source; Own calculations based on World Value Survey
46|————————————————————————————————————————————————————|Palermo Business Review | Nº 7 | 2012
The Economics of Happiness in Argentina
Therefore, in order to get a more appropriate picture of what actually happened from
1984 to 2011, we split the data according to three different Argentinean regions; namely
Buenos Aires City, Great Buenos Aires and the Rest of the Country, and use the sampling
weights within each region to account for socioeconomic differences. Table number 4
shows the average changes in life satisfaction in 1991, 1995 and 2006 in comparison with
1984, by regions within the country.
Table 5 | The evolution of Life satisfaction over time, across different Argentinean
regions (OLS estimation)
Buenos Aires city
Buenos Aires
sorroundings (GBA)
Rest of the country
Education
0,02
0,04
0,17
(t value)
0,37
0,72
2,82**
0,47
-0,11
0,28
2,61**
-0,38
1,13
0,65
-0,61
-0,26
3,53**
-1,94*
-1,33
0,67
0,26
0,76
(t value)
3,39**
0,83
3,97**
Constant
6,41
7,34
6,73
(t value)
38,19**
23,08**
33,43**
0.019
0.023
0.045
Year 1991
(t value)
Year 1995
(t value)
Year 2006
R2
Source; Own calculations based on World Value Survey
(* significant at 5%; ** significant at 1%)
The evolution of Life satisfaction over time can be confirmed with a simple glimpse
at graph number 3.
Graph 3 | “Taking all in how satisfied are you with your life”, by region
Source; Own calculations based on World Value Survey and Gallup- Universidad de
Palermo (bars represent 95% confidence intervals)
Palermo Business Review | Nº 7 | 2012|————————————————————————————————————————————————————| 47
Martin Tetaz
In graph 3, regional averages are grouped by year of survey. The first 95% confidence
interval bar within each year represents always Buenos Aires City, while the second belongs
to Grate Buenos Aires. The last is the Rest of the country interval. Again overlapping
between two different surveys confidence intervals, for the same region, means the absence
of statistically significant differences, whereas the opposite indicates otherwise.
Clearly, Buenos Aires City´s life satisfaction jumped from 1984 to 1991, remains
steady to 2006 and depending on the 2011 wave analyzed, either rose again or stayed
about the same. On the other hand, life satisfaction in the surroundings of Buenos Aires
City (GBA) did not improved in 1991 (from 1984) and actually fell in 1995, recovered
in 2006 and stayed basically the same since then. Finally, average answers in the Rest
of the Country showed a similar path to that described for Grate Buenos Aires (GBA),
although with a higher life satisfaction since 2006.
Those life satisfaction patterns are consistent with the structural changes in the
Argentinean economy during this period of time. The over valuation of the local currency
during the 90’s inflicted a huge damage on the industrial sector of the country, basically
located at the Great Buenos Aires. This sector strongly recovered thanks to the devaluation
in 2002. The Rest of the country, in turn, took advantage of the beneficial terms of trade
of the last 7 years, because the countryside specializes in agricultural production.
3. Determinants of Happiness, the econometric approach
Strictly speaking, the former question produces an ordinal variable therefore we should
estimate our model running an Ordered Probit model that assumes the underlying relation
Yi * = α + ∑ β j X ji + µ i
1)
j
Where “Y* i” is the unobserved life satisfaction of subject “i”; α stands for a constat,
the β j`s are parameters, and “X ji” is a vector of “j” personal characteristics of subject
“i”. As usual “µ i” refers to the white noise error term. Because we can only observe ten
possible discrete realizations of “Yi “ the Ordered Probit model estimates by maximum
likelihood the coefficients that maximize the joint probability of the actual observed
values of “Yi “, conditional on the observed “X j” vector of “j” personal characteristics.
Since the interpretation of the resulting coefficients is not straight forward (i.e. they are
not marginal effects), it may be useful to run a simple ordinary last square regression
(OLS) alongside, and use those coefficients as an approximation of the marginal effects
(simulations can be conducted, should an interest on the exact magnitude of a particular
effect arises).
On the other hand, the latter question presents a harder problem because answers
provide categorical data. Normally we start by stating the probability of a particular
48|————————————————————————————————————————————————————|Palermo Business Review | Nº 7 | 2012
The Economics of Happiness in Argentina
realization of the categorical variable, conditional on the “X ji” observed vector of “j”
personal characteristics of subject “i”, as a linear function of some sort.
Pi
= E (Yi = 1 /
m
X ij ) = β1 + ∑ β j X ij (2)
j =1
Then we do the same for the other possible particular realizations of Yi and estimate
(again by maximum likelihood) the β coefficients so as to maximise the probability of a
particular realization of that dependent variable against a rival realization, called the “base
outcome”. The resulting Multinomial Logistic model produces therefore, coefficients that
can be interpreted as the change in the relative (to the base outcome) likelihood function,
due to the change in the independent variable of interest. Coefficients of one particular
category can also be compared relative to other categories.
4. The results: World Value Survey data
Under Easterlin’s Hypothesis there should not be any effect of income on neither life
satisfaction nor happiness, over time. Effects of income on happiness may yet be present at
particular moment; after all, Easterlin indeed found cross section effects within the United
States and systematically richer people reported higher levels of happiness (perhaps due
to relative income effects). Regretfully, our data bases do not include always an income
variable; however the World Value Survey five waves do indeed have the same question
regarding the satisfaction on household financial situation, providing a proxy for income.
In table 5, we show the evolution of the answers to the question “How satisfied are you
with the financial situation of your household?” Presumptively such satisfaction depends
on the income, among other things, but since we don’t have income variables, controlling
for the level of education and gender, we obtain a pseudo Mincer approximation.
Table 6 | The evolution of the satisfaction of household’s financial situation over time,
across different Argentinean regions (OLS estimation)
All country
Buenos Aires
city
Great Buenos
Aires
Rest of the
country
Education
0,288**
0,233**
0,309**
0,367**
Gender (male)
-0,135*
-0,034
-0,027
-0,245
year1991
-0,149
0,251
-0,827**
-0,438
year1995
-0,461**
-0,006
-0,897**
-1,183**
year1999
0,136
year2006
1,045**
1,084**
0,436
0,661**
_cons
4,932**
4,632**
5,258**
5,465**
Source; Own calculations based on World Value Survey
(* significant at 5%; ** significant at 1%)
Palermo Business Review | Nº 7 | 2012|————————————————————————————————————————————————————| 49
Martin Tetaz
Definitely the satisfaction on household’s financial situation varies heterogeneously
over time, depending on the geographical region. In Buenos Aires City, there seems not
to be any improvement in the ’90 (the 1984 satisfaction is the comparison point), but there
is indeed a significant rise in the last wave (2006). In contrast, its surrounding area, The
Great Buenos Aires, depicts a worsening in financial satisfaction in 1991, lasting until
1995. In the last wave the dependent variable rose, but never fully recovered the levels
of the 1994 wave. The Rest of the country got even much worse in 1995, however not
only recovered completely the 1984 level of satisfaction in 2006 but surpassed it, perhaps
coupling the effect of the devaluation of the exchange rate plus the improvement of the
international terms of trade that favoured mostly the countryside. This is consistent with
the over valuation of the exchange rate during the last decade of the previous century,
because it caused a huge level of industrial unemployment everywhere outside Buenos
Aires (the capital city economy concentrates on services that actually benefited from the
low value of foreign currencies).
Accordingly, had Easterlin been wrong, we should expect a similar patter regarding
life satisfaction. Table 6 presents the evolution of happiness across the five waves of the
World Value Survey.
50|————————————————————————————————————————————————————|Palermo Business Review | Nº 7 | 2012
The Economics of Happiness in Argentina
Table 7 | The evolution of Happiness over time, across different Argentinean regions
(Multinomial Logit estimation)
Buenos Aires city
Buenos Aires
surroundings (GBA)
Rest of the country
Coef
z
Coef
z
Coef
z
very happy
education
0,02
0,17
-0,05
-0,7
0,03
0,43
year1991
0,76**
3,52
1,62**
3,78
0,78
2,87
year1995
0,13
0,55
1,23**
2,9
0,62
2,82
year2006
0,55*
2,31
1,22**
2,84
0,52
2,46
_cons
-1,23
-5,67
-1,67
-3,87
-1,06
-4,74
education
-0,24*
-2,51
-0,2
-2,2
-0,26**
-2,81
year1991
0,02
0,08
0,96*
2,43
0,24
0,74
year1995
-0,24
-0,98
0,25
0,63
0,18
0,73
year2006
0,2
0,8
0,3
0,74
-0,71**
-2,64
-0,49
-2,44
-1,15
-2,84
-0,73
-2,82
-0,37
-1,61
-0,46*
-2,37
-0,2
-0,74
1,27
1,09
1,45
not very happy
_cons
not at all happy
education
year1991
2,11**
3,8
1,34
year1995
0,66
0,88
1,1
1,05
-0,03
-0,03
year2006
1,86**
2,85
0,91
0,84
-0,84
-1,02
-3,39
-5,74
-2,91
-2,75
-3,21
-4,23
_cons
(happiness==quite happy is the base outcome)
Source; Own calculations based on World Value Survey
(* significant at 5%; ** significant at 1%)
The first column of the previous table indicates a widening of happiness variance in
Buenos Aires both in 1991 and 2006. More people feel “very happy” rather than just “quite
happy” in the first wave of the 90´s , but the number of those reporting to be “not at all
happy” rose as well, being the coefficients of the latter category more than threefold the
size of those belonging to the former. On the other hand, in Buenos Aires surroundings,
happiness increased unambiguously in 1995 and stayed stable until 2006. Last but not
least important, fewer people claimed to be “not very happy” in the sample of the 2006
wave, living in the “rest of the country”. From table 4 we also know that life satisfaction
improved in Buenos Aires City during the 90’s, but stayed roughly the same in 2006,
whereas the satisfaction to the financial situation of the household had not shown any
increase neither in 1991 nor in 1995, but a huge rise in 2006.
Palermo Business Review | Nº 7 | 2012|————————————————————————————————————————————————————| 51
Martin Tetaz
As to Great Buenos Aires samples, table 4 indicated a worsening since 1995, with no
reversal in 2006, but from table 5 is clear that financial satisfaction in those households
had dramatically fell earlier, in 1991.
Finally, when it comes to the Rest of the Country data, life satisfaction improved in
2006, following the recovery of the financial satisfaction of this year. Nonetheless, the
former did not show any worsening during the ’90 even when the latter clearly fell in
1995, according to table 5.
Taking all in, Easterlin appears to be vindicated; therefore we devote the rest of the
paper to study the (other) determinants of life satisfaction. The next table presents the
estimations of five different ordinary least squares regressions; one for each wave of the
World Value Survey.
52|————————————————————————————————————————————————————|Palermo Business Review | Nº 7 | 2012
The Economics of Happiness in Argentina
Table 8 | Determinants of Life satisfaction over time (OLS estimation)
Year 1984
Year 1991
Year 1995
Year 1999
Year 2006
Coef
t stat
Coef
t stat
Coef
t stat
Coef
t stat
Coef
t stat
age
-0,06
-2,02
-0,07
-2,32
-0,08
-3,47
-0,05
-1,88
age (square)
0,00
2,17
0,00
2,15
0,00
3,00
0,00
1,75
Buenos Aires City
-0,58
-2,66
-0,45
-1,77
0,22
1,02
-0,71
-4,53
Great Buenos Aires
0,39
1,34
0,00
-0,01
-0,08
-0,47
-0,09
-0,61
Gender (male)
0,02
0,12
0,26
1,53
-0,04
-0,23
0,05
0,35
0,16
1,25
Class A and B
-0,08
-0,26
Class C2
-0,13
-0,65
Class D
-0,96
-3,39
Income Decil
0,07
2,15
Income level
0,00
0,74
Education level
0,16
1,49
-0,10
-0,93
Primary educ incompleted
-0,17
-0,48
-0,10
-0,38
-0,06
-0,20
Primary educ completed
0,55
2,30
-0,16
-0,88
-0,39
-2,03
Secondary educ incompleted 0,20
0,93
0,04
0,25
-0,12
-0,71
Superior educ incompleted
0,21
0,70
0,10
0,38
-0,03
-0,17
Superior educ completed
0,33
1,08
0,41
1,90
0,02
0,08
Single
-0,13
-0,39
-0,33
-1,01
-0,93
-2,97
-0,47
-2,00
-0,35
-1,38
Live together
0,96
2,39
-0,42
-1,01
-0,95
-2,84
-0,13
-0,53
-0,15
-0,77
Separated
-1,53
-3,02
-0,63
-1,59
-1,04
-2,17
-1,09
-3,64
-0,75
-2,27
Divorced
1,00
1,89
-0,10
-0,14
0,21
0,42
-0,31
-0,65
-0,68
-2,07
Widowed
-0,20
-0,75
-0,75
-2,00
-0,46
-1,36
-0,54
-1,67
-0,40
-1,21
One child
0,27
0,79
-0,07
-0,22
0,18
0,63
0,08
0,36
-0,47
-2,09
Two children
0,19
0,64
0,25
0,79
-0,20
-0,68
0,36
1,54
0,04
0,19
Three children
0,25
0,77
-0,32
-0,93
-0,30
-0,99
0,06
0,24
0,12
0,52
Four children
-0,16
-0,31
0,19
0,47
0,06
0,14
0,51
1,81
-0,41
-1,20
Five or more children
-0,28
-0,40
0,04
0,05
0,10
0,16
0,50
1,18
0,13
0,28
_cons
6,68
13,63
8,69
10,22
8,59
10,59
9,16
15,69
9,10
13,83
R2
0,05
0,05
0,05
0,04
0,05
Source; Own calculations based on World Value Survey
(Shadowed coefficients significant at 5%)
Palermo Business Review | Nº 7 | 2012|————————————————————————————————————————————————————| 53
Martin Tetaz
For 1984, we do not have respondents age, neither we have any variable to account for
income, conforming ourselves with just an imputed value of education1. To begin with
the statistically significant results, people living in Buenos Ares City felt less satisfied
with life than those in the Rest o the Country (the omitted dummy variable). Those living
together reported higher levels of satisfaction compared with married ones. Interestingly
the same can be said for divorced subjects. In contrast there seems to be a sadness effect
for separated people. Finally, children do not make any difference.
In 1991 we have respondents age and social class, both playing a significant role. As
mentioned earlier, the “U” shape effect of age on life satisfaction is a commonly result
in the literature, and not only we find the same result here, but the impact is basically the
same across all remaining survey waves, as shown in the graph 4, indicating that age has
a negative impact on life satisfaction until somewhere between 45 and 55 years of age
(depending on the wave analyzed).
Graph 4 | The relation between life satisfaction and age
Source; Own calculations based on World Value Survey
Regarding geographical effects, people in Buenos Aires City remain less satisfied
than the Rest of the Country, although the coefficient is just significant at 10% level. As
to social class, those belonging to the lower step (Class D) claimed not to be as happy
as those in Class C1 (the omitted dummy variable for medium class). Contrary to 1984
wave results, widowed were significantly less happy than married, being 1991 the only
wave in which this result arises.
1. In 1984 survey the only variable with information about education is a question asking respondents the year
they left education. Fortunately that variable correlates highly with level of education in the other waves, allowing
us to estimate education level based on that.
54|————————————————————————————————————————————————————|Palermo Business Review | Nº 7 | 2012
The Economics of Happiness in Argentina
For the 1995 wave, the World Value Survey provides information on income deciles
and the positive and statistically significant coefficient indicates a positive (though small)
relation with life satisfaction. In contrast to the previous waves, there is no evidence of
geographical effects.
The wave also has information on education level. Strange as it may be, those with just
primary education completed reported higher life satisfaction than those with secondary
completed (the omitted dummy variable). Marital effects return although this time single,
separated and those living together were less satisfied than the happily married.
When it comes to the 1999 survey, the wave does not have geographic information
so we cannot see whether Buenos Aires City inhabitants remain pessimistic about their
lives. On the other hand, there is now data on incomes allowing us to analyze the impact
of money on happiness. Not surprisingly, the coefficient is positive but not significant
at all. However, people having completed superior education indeed claim to feel more
satisfied with their lives, compared with those with secondary education (the omitted
dummy variable). In accordance with the precedent wave, both single and separated
subjects felt less satisfied than married (the omitted dummy variable).
Finally, the last wave presents again a negative Buenos Aires City effect and a negative
impact on life satisfaction for those with just complete primary education (relative to
those with secondary). Separated and divorced are less happy than married, and we find
a noteworthy child negative effect.
Summing up, age “U” shape effects are always present, Buenos Aires citizens feel less
happy, money plays a modest role, education effects are neither consistent nor stable, and
separated people are always worse. Last but not least, since table 7 estimations come from
an OLS regression, coefficients are actually marginal effects measuring the change in the
dependent variable (life satisfaction) due to the one unit modification of any particular
independent variable.
5. Satisfaction on the financial situation of the households and endogeneity
Issues
It is important to notice that in order to analyze to what extent money can buy
happiness, we can further exploit the World Value Survey data base, taking advantage of
the already mentioned question that measures the satisfaction on the financial situation of
the household. A potential concern may be that since this information also comes from a
subjective evaluation, it may be endogenous on life satisfaction. To properly address this
issue, in table 8 we show the results of a three stages simultaneous estimation of “Life
satisfaction” and “Financial satisfaction”.
Palermo Business Review | Nº 7 | 2012|————————————————————————————————————————————————————| 55
Martin Tetaz
Table 9 | Determinants of Life satisfaction over time (simultaneous equations, 3 stages
OLS estimation)
Life satisfaction
Year 1984
Year 1991
Year 1995
Year 1999
Year 2006
Coef
Coef
t statt
Coef
t stat
Coef
t stat
Coef
t stat
t stat
Satisfaction financial situation 0,29
1,83
0,62
2,99
0,38
2,17
0,37
2,47
0,44
1,62
Buenos Aires City
-0,45
-2,36
-0,37
-1,78
0,11
0,77
-0,39
-2,38
0,2
0,82
0
0,02
0,02
0,23
-0,06
-0,74
age
0
-0,08
-0,02
-0,73
-0,04
-1,41
-0,03
-1,42
age (square)
0
0,27
0
0,74
0
1,18
0
1,4
Gender (male)
-0,03
-0,28
0,2
1,38
-0,08
-0,76
-0,01
-0,06
0,1
1,28
Single
-0,07
-0,34
-0,19
-0,69
-0,46
-1,99
-0,4
-2,12
-0,23
-1,4
Great Buenos Aires
Live together
0,52
1,09
-0,28
-0,7
-0,52
-1,96
-0,15
-0,96
-0,16
-0,95
Separated
-1,15
-2,44
-0,42
-1,25
-0,86
-2,1
-0,75
-2,7
-0,61
-1,58
Divorced
0,53
0,87
0,02
0,02
0,24
0,71
-0,38
-1,19
-0,37
-1,65
Widowed
0,03
0,16
-0,6
-2
-0,3
-1,42
-0,36
-1,81
-0,25
-1,45
One child
0,17
0,6
0,06
0,17
0,27
1,09
0,1
0,52
-0,25
-1,19
Two children
0,13
0,55
0,29
0,96
-0,14
-0,6
0,31
1,72
-0,02
-0,1
Three children
0,21
0,78
0,06
0,17
-0,22
-0,87
-0,05
-0,23
0,28
1,29
Four children
-0,24
-0,59
0,21
0,5
0,3
0,95
0,34
1,43
-0,19
-0,71
Five or more children
-0,21
-0,37
-0,18
-0,29
0,48
1,11
0,48
1,42
0,35
0,94
Constant
5,44
5,41
3,78
2,31
5,83
4,08
6,3
4,44
5,67
2,67
Satisfaction financial situation Coef
t stat
Coef
t stat
Coef
t stat
Coef
t stat
Coef
t stat
Life satisfaction
0,91
4,38
0,31
1,07
0,94
4,05
0,87
3,68
0,82
3,5
age
-0,07
-1,86
-0,05
-1,43
-0,05
-1,72
0
0,09
age (square)
0
1,8
0
1,37
0
1,86
0
-0,04
Class A and B
0,16
0,52
Class C2
-0,34
-1,85
Class D
-1,08
-2,88
Income Decil
0,08
2,3
Income level
0
2,14
0,41
3,46
-0,02
-0,21
0,07
0,39
0,3
2,04
0,3
1,78
0,08
0,44
0,28
2,01
0,34
1,95
0,37
1,33
0,44
1,7
0,39
1,74
Education level
Secondary educ uncompleted
Secondary educ completed
Superior educ uncompleted
Superior educ completed
One child
0,46
1,77
0,93
2,92
0,45
1,78
0,09
0,34
-0,12
-0,42
-0,25
-0,82
-0,24
-1,06
-0,05
-0,21
Two children
0,03
0,12
-0,06
-0,19
0,19
0,75
-0,33
-1,43
0,11
0,55
Three children
-0,07
-0,25
-0,44
-1,49
0,17
0,62
0,09
0,42
-0,42
-1,72
Four children
0,31
0,69
0,04
0,09
-0,42
-1,1
-0,11
-0,34
-0,1
-0,33
Five or more children
-0,01
-0,01
0,42
0,67
-0,4
-0,73
-0,73
-1,82
-0,61
-1,45
Constant
-1,6
-1,22
4,78
1,88
-1,01
-0,55
-0,2
-0,1
-0,1
-0,05
Source; Own calculations based on World Value Survey
(Shadowed coefficients significant at 5%)
56|————————————————————————————————————————————————————|Palermo Business Review | Nº 7 | 2012
The Economics of Happiness in Argentina
The results on the upper panel suggest a positive and (almost always) highly significant
effect of financial satisfaction on life satisfaction. This result may appears to conflict with the
earlier confirmation regarding the Easterlin effect. Nevertheless it is noteworthy to say that
within a society at any particular time there may exist indeed an effect whereby the richer
feels happier even when across different countries or societies that effect is absent, as shown
by Esterlin. The same happens when we observe the same country over time. Perhaps the
explanation has to do with the relevance of relative income for happiness, over and above
absolute income or wealth. Notwithstanding, however, the coefficients are all below one,
implying that the effect is less than proportional. As to age, it impacts the estimations on the
lower panel, perhaps indicating that older people feel worse through a financial satisfaction
channel. To complete, the other coefficients do not differ too much from those of table 7.
In the next section we study other variables available in the last two 2011 surveys, run
by Gallup and Universidad de Palermo.
6. The Gallup – Universidad de Palermo survey
To begin with, we estimate Life satisfaction first by Ordinary Least Squares (columns
1 and 2) and then by Ordered Probit (columns 3 and 4). The next table depicts the results
from the first wave of the 2011 survey.
Table 10 | Determinants of Life satisfaction
OLS estimation
Ordered Probit estimation
Coeficients
t statistics
Coeficients
Z statistics
Age
-0,044*
-2,130
-0,029*
-2,130
Age (square)
Gender (male)
0,000
-0,114
1,730
-1,010
0,000
-0,107
1,740
-1,460
Class ABC1
0,515*
2,270
Class D1
-0,043
-0,330
0,351
-0,009
1,830
-0,110
Class D2E
-0,449**
-2,660
-0,255**
-2,490
Single
-0,460*
-2,050
-0,295*
-2,140
Live together
0,252
-0,293
1,690
-1,200
0,148
1,420
Separated / divorced
-0,165
-1,080
Widowed
-0,803**
-2,800
-0,454**
-2,740
Children
-0,078
-0,410
-0,037
-0,310
Buenos Aires city
-0,662**
-4,750
-0,480**
-5,460
-0,216**
-2,850
Great Buenos Aires
-0,246*
-2,060
_cons
9,546**
20,000
R2
0,08
0,02
Source; Own calculations based on Gallup-Universidad de Palermo
(* significant at 5%; ** significant at 1%)
Palermo Business Review | Nº 7 | 2012|————————————————————————————————————————————————————| 57
Martin Tetaz
It is clear from columns 1 to 4 that age effects remain stable under any estimation
strategy, confirming the accordance with the international usually found results. Older
subjects are less satisfied although the effect diminishes as people ages. The extremes
of the social classification present a logical pattern whereby high class individuals
(ABC1) feel happier and conversely low class respondents (D2E) are less satisfied
with their lives.
Single and widowed interviewed, report lower levels of life satisfaction (relative to
married). People living in Great Buenos Aires are less satisfied than those in the Rest
of the Country (the omitted dummy variable), but inhabitants of Buenos Aires City are
much less satisfied.
Even when we showed above the fact that life satisfaction and happiness highly
correlate, it may be interesting to explore the effect of some of the previous variables
in a qualitative framework. The following table shows a multinomial logit regression.
Table 11 | Determinants of Happiness (Multinomial Logit estimation)
Not at all happy
Not very happy
Coef
Z stat
Coef
Z stat
Coef
Z stat
age
0,006
0,05
0,067*
1,81
-0,089**
-2,93
0
0,34
0
-1,4
0,001**
2,68
age (square)
Very Happy
Gender (male)
0,044
0,06
0,04
0,19
-0,213
-1,29
Class ABC1
0,661
0,5
-1,249*
-2,16
0,109
0,23
Class D1
0,502
0,44
0,711**
2,8
0,281
1,54
Class D2E
1,716*
1,89
0,768**
2,63
-0,15
-0,67
Single
1,339
1,45
0,62
1,48
-0,488
-1,65
Live together
-12,783**
-15,71
-0,241
-0,66
-0,176
-0,77
Separated / divorced
0,708
0,63
0,034
0,08
0,144
0,47
Widowed
-0,929
-0,88
0,873*
2,37
-1,027**
-2,52
Children
0,597
0,48
-0,189
-0,53
0,106
0,42
-13,389**
-33,05
0,528*
2,04
-0,832**
-3,49
Buenos Aires city
Great Buenos Aires
_cons
-0,302
-0,42
0,074
0,31
-0,167
-0,99
-6,319**
-2,71
-3,861**
-3,87
1,771**
2,53
Source; Own calculations based on Gallup-Universidad de Palermo (* significant at 5%; **
significant at 1%) “Quite Happy” is the comparison group
58|————————————————————————————————————————————————————|Palermo Business Review | Nº 7 | 2012
The Economics of Happiness in Argentina
The table should be read considering that coefficients in column one, for example,
indicate to what extent a particular independent variable affects the chance of belonging
to that column category, in opposition to the omitted category (being quite happy). Having
said that, let’s have a look at the coefficients. Age effects show the same previously found
pattern, though age is not responsible for people feeling not at all happy. Social class has
an impact in the expected way. It is less likely to be “not very happy” if someone belongs
to the high class (ABC1), on the other hand, should a subject comes from a low class
family it is more likely that she feels either “not very happy” or “not at all happy”. What
the social class effect is saying, is that a good position is useful to avoid sadness, rather
than to assure happiness.
Interestingly, living together (relative to the omitted married category) represents
almost a full guarantee against feeling “not at all happy”, while being widowed excludes
anyone from the “very happy” group.
When it comes to geographical effects, it is noteworthy that it is less likely to be very
happy in someone lives in Buenos Aires City but the citizenship fully protects people
from falling in the “not at all happy” group.
In order to confirm our findings for the 2011 sample, we now turn to the second wave
of the Gallup Universidad de Palermo survey. The strategy is the same, table eleven shows
the determinants of life satisfaction, and the following table presents the multinomial
happiness study.
Palermo Business Review | Nº 7 | 2012|————————————————————————————————————————————————————| 59
Martin Tetaz
Table 12 | Determinants of Life satisfaction
OLS estimation
Ordered Probit
estimation
coef
t stats
coef
t stats
-0,031
-1,74
-0,023
-1,81
0
1,64
0
1,75
Gender (male)
-0,131
-1,21
-0,129
-1,69
Class ABC1
-0,043
-0,19
-0,05
-0,33
Class D1
-0,023
-0,21
0,013
0,17
Class D2E
-0,129
-0,84
-0,068
-0,64
0,775**
3,58
0,528**
4,11
Active in labour
-0,084
-0,64
-0,069
-0,76
Active religeously
0,138
1,15
0,107
1,25
Active voluntariate
-0,257
-1,76
-0,165
-1,61
Active in sports
0,012
0,11
0,014
0,18
0,426**
2,41
0,227*
1,98
Active in studying
0,111
1,05
0,077
1
Active going out
0,311**
2,67
0,197*
2,38
Active politically
-0,242
-1,31
-0,171
-1,43
Active in couple life but seldom sex
0,442**
2,7
0,298**
2,65
Active in sex life but not at couple
0,288
1,23
0,216
1,25
Active in both couple and sex lives
0,555**
3,97
0,391**
3,99
Buenos Aires city
-0,2
-1,31
-0,115
-1,13
Great Buenos Aires
-0,1
-0,84
-0,09
-1,07
7,231
15,54
Age
Age (square)
Active in family life
Active socially
_cons
R2
0,12
0,04
Source; Own calculations based on Gallup-Universidad de Palermo
(* significant at 5%; ** significant at 1%)
60|————————————————————————————————————————————————————|Palermo Business Review | Nº 7 | 2012
The Economics of Happiness in Argentina
We find again, the same previous age effect, although this time is less statistically
significant (just at 10%). Contrary to the first wave, there is no statistically significant
social class effect, and neither have we found geographical regions effects.
On the other hand, this particular survey asked people how active they were in
several domains, providing us with very rich information. For simplification we grouped
responses in a dichotomy variable whereby those either “very active” or “quite active”
were considered “Active”. The results are very interesting, indicating that those active
in their family life, socially active and accustomed to going out frequently, felt more
satisfied with their lives.
Moreover, we split dating and sexual activity levels in four different groups; those
active sexually but not very active in other aspects of their romantic relationships,
those active in their couple lives, but not very active sexually, a third class containing
people active both in their relationships and sexually, and a fourth group of those neither
active sexually nor in any other aspect of a romantic relationships (this particular group
was the omitted variable, so results should be interpreted as relative to be in this class
of people). As expected, best case scenario was being active in dating someone and
having sex frequently (0,55 more satisfied with their lives), followed by just dating
someone actively, though without a very active sexual life (0,44 additional points in
life satisfaction). Perhaps surprisingly, people very active in their sexual lives, but not
active romantically, did not report higher levels of life satisfaction than those neither
dating nor having sex.
Finally, in the same spirit we did it above, we now turn to a multinomial analysis
of happiness. Please bear in mind that coefficients in column one of the next table, for
example, indicate to what extent a particular independent variable affects the chance of
belonging to that column category, in opposition to the omitted category (being quite
happy).
Palermo Business Review | Nº 7 | 2012|————————————————————————————————————————————————————| 61
Martin Tetaz
Table 13 | Determinants of Happiness (Multinomial Logit estimation)
Not at all happy
Coef
Z stat
Age
0,375
1,81
Age (square)
-0,004
-1,8
Gender (male)
-4,151**
-3,23
Class ABC1
-17,878**
-8,87
Not very happy
Coef
Very HappY
Z stat
Coef
Z stat
0,100*
2,14
-0,033
-1,18
-0,001*
-2,18
0
0,8
-0,105
-0,38
-0,147
-0,86
-2,444*
-2,28
-0,095
-0,24
Class D1
2,222*
2,26
0,484
1,57
0,215
1,18
Class D2E
1,686
1,57
1,089**
3,07
0,496*
2,12
Active in family life
2,632*
2,35
1,008**
2,92
0,564**
2,6
Active in labour
0,094
0,1
-0,02
-0,06
-0,358
-1,9
Active religeously
-3,425**
-3,46
-0,239
-0,72
1,778**
3,96
Active voluntariate
1,244
1,27
-0,950**
-3,01
-0,506**
-2,58
Active in sports
1,79
1,49
0,062
0,21
0,08
0,43
Active socially
0,369
0,38
0,366
0,94
-0,062
-0,28
Active in studying
-12,617**
-9,72
0,121
0,37
0,149
0,84
Active going out
-0,906
-1,41
-0,951**
-3,23
0,248
0,92
Active politically
-1,153
-1,33
0,469
1,52
0,468**
2,64
Active in couple life but
seldom sex
-3,498
-1,65
-0,694*
-2,24
-0,187
-1,01
3,396**
2,62
-0,174
-0,33
-0,426
-1,52
Active in both couple and
sex lives
-2,25
-1,6
-0,17
-0,46
0,498
1,91
Buenos Aires city
1,655
0,79
0,557
1,23
0,756
1,84
Great Buenos Aires
-1,93
-1,28
-0,477
-1,32
0,747**
3,17
-11,287**
-2,21
-2,701*
-2,41
-2,066**
-2,87
Active in sex life but
not at couple
_cons
Source; Own calculations based on Gallup-Universidad de Palermo
(* significant at 5%; ** significant at 1%)
Over and over again, older people are significantly (though just at 5% and 10% levels)
less happy. Gender plays a role for the first time. The negative coefficient in column one
says that males are less likely to belong to the “not at all happy “group. High class subject’s
probability of being “quite happy” is significantly higher than the likelihood of belonging
to the “not very happy” and “not at all happy” categories. Medium low class coefficient
is positive in the “not at all happy” column suggesting this class is overrepresented in
the low happiness group. An interesting result arises in low class individuals. There is
a lower likelihood of belonging to the “quite happy” category, relative to feeling either
worse (not very happy) or better (very happy), though the former state is twice likely
than the latter.
62|————————————————————————————————————————————————————|Palermo Business Review | Nº 7 | 2012
The Economics of Happiness in Argentina
The same puzzling result is found when it comes to people active in their family
life. Moreover, active people overpopulate the “not at all happy” group more than other
categories. In addition we find a noteworthy religious effect. Those active in their religious
lives escape the “not at all happy” group and overcrowd the “very happy” category.
Studding seems to be another vaccine against extreme sadness. People active in their
romantic relationships but not as active in their sexual lives, belong mostly to the “quite
happy” class, whereas those with the opposite pattern seems to be condemned to sadness.
Again, heaven is associated with people active both romantically and sexually.
7. Discussion and conclusions
To our knowledge this is the first paper on the economics of happiness using exclusively
data from Argentina. Consequently several results are certainly novel, starting with the
confirmation of the positive and regionally heterogeneous evolution of happiness over time.
Argentineans are happier now than in 1984, but the highest improvement was documented
in Buenos Aires City, being that the only region with a monotonic rise over time. Both
Great Buenos Aires and the rest of the country fell in 1995, and recovered latter.
In accordance with Easteling seminal paper, the improvements in happiness did not
correlate with the satisfaction with the financial situation of the households. Financial
satisfaction did not rise in Buenos Aires city until the 2006 survey, fell in Great Buenos
Aires in 1991, stayed the same during the 90s, but not fully recovered in 2006. Only for
the Rest of the Country sample, data seems to support the hypothesis of a positive relation
between income and life satisfaction.
In addition, analyzing one particular wave at a time, we found consistent effects in
age, indicating that older people fell less satisfied, although the marginal effect diminishes
as people ages. Seldom have we found any gender effect, but once present females were
happier. Social class status and income affects happiness in many but not in all waves. The
sign was almost always the expected. Perceived satisfaction with the financial situation
of the household indeed affects life satisfaction, but it does it less than proportionally.
Moreover, separated couples were less satisfied with life in comparison with married
ones. Divorce effect was ambiguous over time (positive in the first wave, but negative
in the last) perhaps showing the effect of changes in law that facilitated the process. The
same happened for people living together in 1984 (positive effect) and in 1995 (negative).
Children coefficients were almost never statistically significant, and when they were,
actually having one child was found to be detrimental for life satisfaction (in the 2006
wave). People active socially, that frequently go out and have a fluid familiar relationship,
were more satisfied with their lives. Particularly, religious activity and studying were
fundamental to escaping extreme sadness. Finally, being active in a relationship is
important, but coupling that with a lot of sex is paramount, even when the sex itself does
not make any difference whatsoever.
Palermo Business Review | Nº 7 | 2012|————————————————————————————————————————————————————| 63
Martin Tetaz
Bibliography
Aknin, Lara B., Elizabeth W. Dunn, and Michael I. Norton. (2012) “Happiness runs in a
circular motion: evidence for a positive feedback loop between prosocial spending and
happiness”. Journal of Happiness Studies 13: 347-355.
Blanchflower D. Oswald A. (2004). “Well-being over time in Britain and the USA”.
Journal of Public Economics, 88, 1359-1386.
Clark A. FrijtersP. Shields M. (2008). “Relative Income,Happiness and Utility: An
Explanation for the Easterlin Paradox and Other Puzzles.” Journal of Economic Literature
46, no. 1: 95–144.
Cruces G. Ham A. Tetaz M. (2008). “Quality of Life in Buenos Aires’ Neighborhoods:
Hedonic Price Regressions and the Life Satisfaction Approach”, Working paper for the
project “Quality of Life Urban Neighborhoods in LAC”, BID, Research department, 2008.
Easterlin R. (1974). “Does Economic Growth Improve the Human Lot? Some Empirical
Evidence.” In Nations and Households in Economic Growth: Essays in Honor of Moses
Abramowitz, edited by Paul A. David and Melvin W. Reder. Academic Press.
Frey S.Stutzer A.(2002). “What Can Economists Learn from Happiness Research?”
Journal of Economic Literature 40, no. 2: 402–35.
Giarrizzo, V.(2008). “Economía y Felicidad: Existe vínculo?”. Unpublished.
Graham C. (2008). “Happiness and Health: Lessons—And Questions—For Public Policy.”
Health Affairs 27, no. 1: 72–87.
Powdthaveel N. (2008). “Economics of Happiness: A Review of Literature and
Applications”. Forthcoming in Chulalongkorn Journal of Economics.
Powdthavee, N. (2010) “How Much Does Money Really Matter? Estimating the Causal
Effects of Income on Happiness”, Empirical Economics, 39(1), 77-92.
Powdthavee, N. (forthcoming) “Is Personality Fixed? Personality Changes as Much as
“Variable” Economic Factors and More Strongly Predicts Changes to Life Satisfaction”,
with Christopher Boyce and Alex Wood, Social Indicators Research.
Ravallion M. Lokshin M. (2002). “Self-rated economic welfare in Russia”. European
Economic Review, 46, 1453-1473.
Stevenson B. Wolfers J. (2008). “Economic Growth and Subjective Well Being; reassessing
the Easterlin Paradox”. Brookings Papers on Economic Activity.
Van Praag B. Baarsma B. (2004). “Using happiness surveys to value intangibles: The
case of airport noise”. Economic Journal, 115, 224-246.
64|————————————————————————————————————————————————————|Palermo Business Review | Nº 7 | 2012
The Economics of Happiness in Argentina
Veenhoven R. Hagerty M. (2006). “Rising Happiness in Nations 1946-2004; a replay to
Easterlin”.
Social Indicators Research, Vol. 79, pp 421-436.
WORLD VALUES SURVEY 2005 OFFICIAL DATA FILE v.20090621, 2009. World Values
Survey Association (www.worldvaluessurvey.org). Aggregate File Producer: ASEP/JDS,
Madrid.
Palermo Business Review | Nº 7 | 2012|————————————————————————————————————————————————————| 65